论文部分内容阅读
由于高斯代理模型精度易受数据质量和数量的限制且模型固定后无法随着算法进展自适应调整,而近邻法在算法初期因样本数限制导致算法无法达到所需的预测精度,因此,针对优化函数未知的区间多目标优化问题,提出一种融合高斯建模和近邻法求解种群个体支配关系的NSGA-II算法.该算法通过高斯过程对训练样本集进行建模,利用遗传算法对代理模型进行超参数求解,进而通过高斯代理模型得到待测解之间的可能度概率;利用近邻法对待测解和样本解进行相似性计算,得到待测解之间的可能度概率;通过渐消记忆动态调整高斯过程和近邻法所得支配性结果在算法中所占的比重,得到种群个体之间的支配关系.仿真结果验证了所设计算法的有效性.
Because the accuracy of Gaussian proxy model is limited by the quality and quantity of data, and the model can not be adaptively adjusted as the model is fixed, the nearest neighbor method can not achieve the required prediction accuracy due to the limitation of the number of samples in the initial stage of the algorithm. Therefore, This paper proposes an NSGA-II algorithm to solve the problem of multi-objective optimization in interval-unknown functions by using Gaussian modeling and nearest neighbor method to solve the dominance relationship among populations. This algorithm uses Gaussian process to model training samples and uses genetic algorithm to perform proxy model Then the probability of probability between the solutions to be measured is obtained by using the Gaussian proxy model. The neighbor probability method is used to calculate the similarity between the solution to be tested and the sample solution to get the probabilities of probabilities between the solutions to be measured. By adjusting the proportion of the dominant results obtained by the Gaussian process and the nearest neighbor method in the algorithm, the dominance relationship among individuals in the population is obtained. Simulation results verify the effectiveness of the proposed algorithm.